Anthropic Launches Claude Duet: A Proactive AI Agent That Watches Your Work and Asks Before It Acts
After Google's Gemini agent, Anthropic has now joined the race for proactive AI agents. On October 21, 2026, Anthropic announced Claude Duet, a "proactive" AI agent for professionals and small businesses. Instead of waiting for a prompt like a typical chatbot, Duet watches your connected apps, flags what needs attention and handles assigned tasks in the background. It is launching as a research preview behind a waitlist.
Duet follows Meta's Muse and OpenAI's Dots, two personal assistants built around the same proactive idea. The competition is no longer about how well AI answers questions, but about reliability, control and safety once agents touch real data.
What Claude Duet does
Morning brief: Duet sends a short daily rundown of what matters across connected accounts such as email, calendar and work tools.
Regular check-ins: every few hours it reviews your apps and only messages you when something actually needs you.
Its own cloud computer: the agent runs on a virtual machine with its own browser, so it keeps working after you close the app or switch off your device.
Broad connectivity: Duet uses connectors and plug-ins from the Claude Marketplace, which TechCrunch puts at more than 6,000.
Model: Duet runs on Claude Opus 5.5, Anthropic's most capable model today.
It asks before it acts
Anthropic's biggest emphasis is approvals. Duet can read, draft and prepare work on its own, but it waits for your sign-off before sensitive actions such as sending messages, making payments, signing documents or deleting files. According to TechCrunch, Duet can make online purchases through Stripe Link or Shopify Shop Pay, and those purchases also require confirmation. When given access to accounts, Duet does not see passwords or login details.
Anthropic also acknowledges that the agent "can still misjudge what's urgent," which is part of why it remains a preview.
Who can use it and what it costs
For now it is limited to Claude Pro and Max subscribers in the US, who join a waitlist and are admitted in stages.
It is available through a standalone Duet mobile app or on the web at claude.ai.
There is no extra charge beyond the existing plans, but background tasks and account checks count against your plan's usage limits, so heavy use can drain your allowance faster.
Team and Enterprise plans are not supported yet, and Anthropic has not announced a date for a wider rollout.
The Agenorix view: proactive agents need clear boundaries
Duet is not available in Vietnam yet, but the "proactive agent plus human approval" model will soon become the norm. When we build AI chatbots and AI agents that connect to existing systems, we always put three things in place first:
Tidy data and tools: agents are only useful when email, calendar, CRM and documents are well organized with specific access rights.
Approval boundaries: list what an agent may do on its own (read, summarize, draft) and what always needs a human (sending to customers, spending money, deleting data).
Usage cost tracking: background agents consume allowance continuously, so measure before rolling out to a whole team.
Where to start?
Pick one task you repeat every day, such as a morning email digest, appointment reminders or collecting customer requests, and write down which parts an agent may handle alone and which need your approval. That blueprint carries over to Duet or any agent that comes next. If you want a "Duet" tailored to your own workflows, the Agenorix team is ready to help, and your first consultation is free: get in touch.
Sources: CNBC, TechCrunch, MacRumors, 9to5Mac, CBS News (October 2026).
Share